Re: PSPP back in Debian

[Ben Pfaff]

Jeremy Lavergne <address@hidden> writes:

> Just a random heads up: Apple released a new set of dev tools,
> so I've rebuilt everything today using ge5f6c5. That being
> said, I get two tests failing now.
>
> 416 and 420 (memory exhausted). PSPP has sucked up over 1.31GB
> of RAM (and counting) just for 416. I gave the VM only 1.5GB
> which I should think is plenty. If you recommend I give it
> more, let me know.

Ouch.  Those are the following tests?

    416. MEANS standard errors (means.at:201): ok     (0m0.020s 0m0.000s)
    420. MEANS user missing values (means.at:459): ok     (0m0.020s 0m0.004s)

They should not use a significant amount of memory (not if they
complete in 20 ms each).  John, you wrote MEANS, is there any
MEANS-specific reason that these tests could go haywire?

Jeremy, can you try starting one of these in a debugger, letting
it go for a second or two, then getting a backtrace?  The first
one simply consists of running the following PSPP syntax:

----------------------------------------------------------------------
set format F12.4.
data list notable list /id * group * test1 *
begin data.
1 1 85
2 1 90
3 1 82
4 1 75
5 1 99
6 1 70
7 2 66
8 2 52
9 2 71
10 2 50
end data.

means test1 by group 
        /cells = mean count semean seskew sekurt.
----------------------------------------------------------------------

Thanks,

Ben.
-- 
Ben Pfaff 
http://benpfaff.org